odeon application note calibration of impulse response … · theory the diffuse field calibration...
TRANSCRIPT
Application note – Diffuse Field Calibration
1
ODEON APPLICATION NOTE
Calibration of Impulse Response Measurements
Part 1 – Diffuse Field Method
GK, CLC - November 2013
Scope In this application note we explain how to use the Diffuse-field calibration tool in ODEON in order to
derive level-depending room acoustic parameters, such as Sound Strength (G) or STI. It should be
emphasized that all other parameters provided by ODEON do not require calibration since they depend
only on the shape of impulse response and the relative energy decay (see more about the measuring
system and calibration in the ODEON 13 manual[1]).
The calibration in this application note
was done using the Large Reverberation
Chamber at the Technical University of
Denmark. The volume of the chamber is
245 m3 and provides a reverberation
time of about 8 sec at low frequencies.
Reflective panels are placed near the
walls at various angles in order to
distribute reflections evenly and
enhance the diffusivity of the sound field.
Even when another type of room is used
for the calibration, a diffuse sound field
should be achieved, utilizing some
scattering surfaces or objects. A perfect,
rectangular room with hard walls would
lead to flutter echoes (visible at the
impulse response) that would violate the
assumptions of linear decay, taken for
the diffuse-field calibration.
Equipment A highly reverberant and diffuse field room is needed for the calibration. The ideal type of room is a
reverberation chamber. If such is not available, a room with hard walls could be acceptable (eg. a big
bathroom). However the chosen room should be as diffuse as possible meaning that highly symmetric
shapes (eg. rectangular) must be avoided and some extra treatment might be required: placing of hard,
scattering objects around the walls or the floor. It is a good practice to check always for the degree of
linearity of the decay curve in each octave band through one of the XI parameters. If XI exceeds 10 o/oo
this is an indication that the decay is strongly non-linear, because of flutter echo or due to impulsive noise
during the measurement. You can read more about the XI parameter in the ODEON help file, which can be
Application note – Diffuse Field Calibration
2
easily accessed by pressing F1 while working with the measuring system. Apart from this requirement, the
hardware needed for calibration is exactly the same as for the sweep-signal measurement:
1) An omni-directional loudspeaker. 2) An omni-directional microphone. 3) Amplifiers for the speaker (if not an active one is used) and microphone. 4) Audio interface. 5) Lap-top PC. 6) ODEON 12, any edition. 7) Ear protectors.
Theory The diffuse field calibration is explained in the ISO standard 3382-1, for performance spaces[2]. By
definition the sound strength 𝐺 of an omni-directional source, for a specific frequency, is given as the
logarithmic ratio of the sound energy (squared and integrated sound pressure) of the measured impulse
response to that of the impulse response measured in a free field at 10 m distance from the source:
𝐺 = 10log10
∫ 𝑝2(𝑡)𝑑𝑡∞
0
∫ 𝑝102 (𝑡)𝑑𝑡
∞
0
dB (1)
or
𝐺 = 𝐿𝑝𝐸 − 𝐿𝑝𝐸,10 dB (2)
where 𝐿𝑝𝐸 = 10log10 [1
𝑇0∫
𝑝2 (𝑡)𝑑𝑡
𝑝02
∞
0] and 𝐿𝑝𝐸,10 = 10log10 [
1
𝑇0∫
𝑝102 (𝑡)𝑑𝑡
𝑝02
∞
0] is the sound pressure
exposure level of 𝑝(𝑡) and 𝑝10(𝑡) respectively.
The variables in these equations are as follows:
𝑝(𝑡): Instantaneous sound pressure of the impulse response at the receiver’s position. 𝑝10(𝑡): Instantaneous sound pressure of the impulse response at 10 meters distance from the source
in free field. 𝑝0: Reference sound pressure, 20μPa. 𝑇0: 1 sec.
The value of 𝐿𝑝𝐸,10 can be evaluated inside a reverberation chamber, using the diffuse field
approximation:
𝐿𝑝𝐸,10 = 10log10𝐴
𝑆0− 37 + 𝐿𝑝𝐸
𝑅𝑒𝑣𝐶ℎ dB (3)
Application note – Diffuse Field Calibration
3
Where 𝐴 is the equivalent absorption area in the room that is calculated from the Sabine’s formula
(diffuse field assumption): 𝐴 = 0.16𝑉/𝑇, with 𝑉 being the volume (m2) and 𝑇 the reverberation time
(sec). In addition, 𝑆0 = 1 m2.
𝐿𝑝𝐸𝑅𝑒𝑣𝐶ℎ is the spatial- averaged sound pressure exposure level as measured in the reverberation chamber.
If we subtract both sides of the equation from the term 𝐿𝑝𝐸 we get:
𝐿𝑝𝐸 − 𝐿𝑝𝐸,10 = −10log10𝐴
𝑆0+ 37 + 𝐿𝑝𝐸 − 𝐿𝑝𝐸
𝑅𝑒𝑣𝐶ℎ dB
which is equivalent to
𝐺 = 37 − 10log10𝐴
𝑆0+ 𝐿𝑝𝐸 − 𝐿𝑝𝐸
𝑅𝑒𝑣𝐶ℎ dB (4)
So 𝐺 can be calculated just by knowing the relative sound pressure levels at the reverberation chamber
𝐿𝑝𝐸𝑅𝑒𝑣𝐶ℎ and the room under measurement, 𝐿𝑝𝐸. The term relative is used to emphasise that no absolute
sound pressure levels are needed. In other words, no calibration with a reference dB level is required,
since our equipment cannot replace the need of a calibrated sound level meter if absolute sound pressure
measurements are needed.
Equation (3) is derived under the assumption that the power level 𝐿𝑤 of an unknown source at a specific
frequency can be calculated from the measured average sound pressure level 𝐿𝑝𝐸𝑅𝑒𝑣𝐶ℎ inside a
reverberation chamber:
𝐿𝑤 = 𝐿𝑝𝐸𝑅𝑒𝑣𝐶ℎ − 6 + 10log10
𝐴
𝑆0 dB(re 10-12 Watt) (5)
The sound pressure level 10 meters away from this unknown source in free field conditions is calculated
according to the spherical spread law:
𝐿𝑝𝐸,10 = 𝐿𝑤 − 11 − 10log10(10𝑚)2
= 𝐿𝑤 − 31 dB (6)
Substituting equation (5) to (6) leads to equation (3). If the sound source had a known power level of 31
dB then the sound pressure level at 10 m would be 0 dB. By choosing the G- ISO 3382-1 spectrum in the
measurement setup (see next section) ODEON derives the sound pressure level for a receiver as if the
source was an omni-directional source of power level 31 dB/Octave band.
NOTE: ODEON does not display the 𝐺 value explicitly in measured or simulated results. It displays the SPL
(Sound Pressure Level) of a source. The value of SPL becomes equal to the value of 𝐺 only when an omni-
directional source type and a power of 31 dB/Octave band is selected in the Measurement Setup (for
measurements) or in the Point Source Editor (for simulations).
Application note – Diffuse Field Calibration
4
Alternatively a Speech ISO 3382-3 spectrum can be chosen in the measurement setup. In this case ODEON
derives the sound pressure level as if the power spectrum of the source was equivalent to a human
speaker. For derivation of STI this type of source should be chosen.
Method Inside the reverberant room we specify a group of source-receiver positions. A good number is 2 source
and 3 receiver positions, giving 6 combinations[3]. We start ODEON 12 and then open the Options>Program
Setup>Measurement Setup window (Figure 1). In this window we should specify:
1) The type of sweep signal to be used - the default Exponential Sweep setting is recommended for calibration. 2) Audio devices – we should select the input (eg. microphone or line in) and output (speaker) devices used for the measurement. 3) Receiver model – keep the default 1 channel: Omni. 4) The source power spectrum can be specified there for used when loading the impulse response after measurement. For calibration of G the G- ISO 3382-1 spectrum should be chosen, while for STI calibration the Speech ISO 3382-3 spectrum should be used.
Once all connections have been fixed we can test the measuring signal by choosing Tools>Measure IR (Sinusoidal
Sweep) or by clicking on the button (Figure 2). A message like: “WARNING! Calibration file not consistent
with selected audio devices. No calibration applied” may be displayed at the bottom of the window but it
can be ignored at this phase. In this window the sweep length and the estimated length of the impulse
response can be adjusted. The longer is the sweep the higher is the suppression of the background noise
in the room. A 3 dB suppression is achieved for a doubling of the sweep length. The Play Test Signal button
plays the sweep signal without recording any measurement. The slider in the middle of the screen adjusts
the internal gain. Any other gain adjustments in the measuring equipment (windows mixer, audio
interface, amplifiers) belong to external gains.
Application note – Diffuse Field Calibration
5
Figure 1: In the measurement setup important settings, like input-output devices can be performed.
Adjusting the levels
The most important requirement for a calibrated measurement is that all external level/gain adjustments
must be set to fixed values during the calibration and the measurement. Only the internal gain can be
freely changed during the measurement, since ODEON itself compensates for the adjustment. So it is
crucial at this point to decide which should be the value of the external gains during the measurement in
order to drive the room with an adequate signal to noise ratio without overloading. When overloading
occurs ODEON automatically cancels the measurement.
Application note – Diffuse Field Calibration
6
Figure 2: Details for the sweep signal can be set in the Measure Impulse Response window.
Figure 3: Levels and gains in the windows mixer are considered external adjustments and should remain fixed during calibration and measurements. A straightforward recommendation is to set all parameters at their maximum value.
Application note – Diffuse Field Calibration
7
The most straightforward way for adjusting the external gains is to maximize all values: Windows volume
can be set to 100% and amplifier knobs can be turned to the maximum value (Figure 3). Then all settings
are easy to remember/replicate during the measurement). However, one should be careful not to
overload and damage the speaker! In addition the gain of the microphone should be set to a level where
the internal noise does not become very high.
Often the amplifiers used are rather powerful for the speakers used so that a much lower gain is sufficient.
The corresponding value must then be clearly noted down for reference. Apart from that, no extreme
sound power output is needed if the equipment is able to derive a sufficient signal to noise ratio (SPL/Noise
parameter inside ODEON) or decay range, in normal room conditions (with moderate ambient noise).
The Internal gain can be changed during the measurement but it should remain stable during the
calibration at a value of convenience. Again a straightforward choice is to place it at the maximum
possible value before leading to overloading or clipping. Then the value can be reduced in the room under
measurement as long as the signal to noise ratio is adequate.
Recording
After fixing the external level adjustments, recordings of impulse responses in the reverberant room can
be performed by clicking the Measure button in the Measure Impulse Response window. An introductory video on
how to obtain measurements in room can be found at [4]. The recordings are saved as .WAV files. Some
trial measurements should be performed initially by varying the internal gain until an adequate signal to
noise ratio is achieved. It can be desirable to make the measurement with the highest possible SNR,
achievable without overload, but this is not required. NOTE: Use ear protectors to prevent hearing
damage! An indication of the quality of the signal to noise ratio can be obtained by loading the recording
on the Load impulse response tool, (under the Tools menu) and checking whether there are any “*” characters
on the calculated room acoustic parameters. A “*” character means that ODEON was unable to derive the
corresponding value from the impulse response (eg. due to insufficient signal to noise ratio/ decay range).
An example of a successful impulse response in the reverberation chamber is given in Figure 4, where all
parameters have been derived. The sound strength parameter (𝐺 value) is denoted as SPL (dB) and is not
calibrated yet.
Deriving the Calibration File
Having a series of impulse responses in the room we can derive the calibration file to be used with the
actual measurement by clicking on Tools>Calibrate Measurements>Diffuse field>One step (requires fixed signal chain).. ODEON
asks for the volume of the reverberant room and then for the impulse responses. The files should be
selected all together using the SHIFT or CTRL key and opened at once. ODEON then derives the calibration
values for the eight octave bands between 63 Hz and 8000 Hz as an average from the different positions
(Figure 5) and asks for a name of the calibration file. The user is prompted to set the file as the active
calibration file in the Measurement Setup, which is going to be used with the measurements. It is possible to
change the calibration file used in the Measurement Setup in the Calibration File menu. However, if the input and
output devices are different from those used during the generation of the calibration file ODEON gives a
Application note – Diffuse Field Calibration
8
warning for device mismatch, meaning that all level depended parameters will not be calibrated. You can
now launch the Measurement Setup window to check whether the active calibration file is the correct one.
Figure 4: Room acoustic parameters from a healthy impulse response recording inside the reverberation chamber. A sufficient decay range for each octave band makes sure that all parameters are derived.
Calibrated Measurements
Measurements can be carried out with the Measure IR (Sinusoidal Sweep) tool or an external stimulus using the
active calibration file in the Measurement Setup. Calibration will be applied on the recorded .WAV files during
processing at the Load impulse response tool, (under the Tools menu). If there is a mismatch between the
input/output devices used for the calibration file and the ones used for the measurement, a warning is
displayed by ODEON and no calibration is applied.
In the ODEON measuring system the chain of calibration and measurement can be reversed: Someone
could measure first and calibrate the equipment afterwards as long as the external gains are fixed. When
the measurement is finished a calibration file can by assigned by clicking Tools>Measurement calibration>Assign
Calibration to Existing Measurements.
An example of calibrated measurement is shown in figure 6. The calibration file from figure 5 was used in
this case. The SPL values correspond to the sound strength of the source since, a G-ISO 3382-1 source
power spectrum has been chosen in the Measurement Setup, as shown in figure 1.
Application note – Diffuse Field Calibration
9
Figure 5: Six impulse responses and calibration values derived by ODEON.
Figure 6: SPL measurement calibrated with the calibration file derived in figure 5. These SPL values correspond to the sound strength G of the source, since a G-ISO 3382-1 sound power spectrum has been chosen in the Measurement Setup (see figure 1).
Application note – Diffuse Field Calibration
10
Conclusion The note describes the steps required for a calibrated measurement of G and STI using a reverberation
(diffuse) room. Another way for calibration is utilizing an anechoic chamber – performing a so called free
field calibration. The sound field for both environments is considered to be fully known, as long as specific
assumptions hold (diffuse field and free field respectively). For this reason the sound power of an
unknown source can be specified from analytic expressions like eq.(5) and calibration of the equipment
towards a known source like G-ISO 3382-1 and Speech-ISO 3382-3 can be done.
The basic steps described in this note hold for a free-field calibration as well and the main difference is
the way the power of the unknown source is derived. For that matter, one could follow the same process
for such a calibration just by selecting Tools>Calibrate Measurements>Free field>One step (requires fixed signal chain).
References
1. C.L. Christensen & G. Koutsouris. ODEON Room Acoustics Software, manual, version 13, Odeon A/S, Denmark 2015 (http://www.odeon.dk/download/Version13/ODEONManual.pdf).
2. ISO standard 3382-1, Acoustics - Measurement of Room Acoustic Parameters – Part 1: Performance Places.
3. ISO standard 3382-2, Acoustics – Measurement of Room Acoustic Parameters – Part 2: Reverberation Time in Ordinary Rooms.
4. Claus Lynge Christensen. Impulse Response Measurements, ODEON video tutorials: http://www.odeon.dk/impulse-response-measurements.